This paper proposes a new framework for accurate detection of the QRS complex based on centered Cumulative Sums of Squares (CUSUM). This detection of the QRS complex is very important for many health applications. To accurately detect the QRS wave, the proposed framework first uses filters to reduce the frequency of big T and P waves. Then, it applies the CUSUM and differential operation to produce a set of event vectors and uses a fixed threshold to detect the QRS candidates. Then, a width feature is used to discriminate the non-QRS complex from the QRS complex event. For the purpose of evaluation, the standard ECG MIT-BIH database has been used. The evaluation results show that our method has an accuracy of 99.66%.